Our guest today is Lukas Fraser, currently the Program Manager for NV5 Geospatial (formerly Quantum Geospatial). Lukas got his start by studying geodesy and geomatics engineering at the University of New Brunswick. Originally, he wanted to get into land survey, but like many students, he ended up taking the first job he was offered, in this case, a job with a local aerial mapping company. This seems to have worked out, as he spent 3 years working for them, before engaging in 3.5 years (and counting) with NV5 on their unmanned systems program.
Drones are cool. One of the cool things about drones, is they do not complain about carrying things. Granted, you probably will not get expected results if you strap your cat with a GoPro onto your DJI Mavic, there will be many complaints there. Drones alone, however, are more than willing to carry a variety of sensors and cameras. As technology has progressed, the options available have expanded. Sensors that previously could only be carried by helicopters, planes, or carted around on the ground, have been scaled down to work with drones, and can still be removed and utilized with another aircraft if desired.
The sensor in particular that we are interested in here, is the LiDAR scanner. LiDAR is light detection and ranging.
The basic concept is that eye-safe near infrared light is emitted at the ground from the sensor at regular intervals, bouncing off whatever it hits.
The sensor then receives back that information. It knows the speed of light, the two-way distance travelled, and the interval it was collected at. Combined with some other position information collected via the IMU and GNSS, you can now create a point cloud based on those different returns.
But wait- what are the IMU and GNSS?
IMU is the Inertial Measurement Unit, it collects information about the drones position such as pitch, roll, and yaw. The GNSS is the Global Navigation Satellite System, which collects the three degrees of freedom of the drone- XY and Z values.
Knowing this information is important for allowing the drone to automatically tie in spatial information to the data being collected. Secondly, knowing this information allows the LiDAR scanner to correct for slight variations from the flight path that it incurs, adding to the overall accuracy.
We can have our LiDAR sensor, but we can also add other sensors and cameras in order to build better conext for our study area. The most popular pairing with LiDAR collection is to also collect ortho images. 2D products, such as orthomosaics, can be combined with the surface or terrain model resulting from 3D products in order to create a more cohesive representation of the target.
Drones generate a lot of data. So much data, that the software intended to process it needs for it to be thinned before it can be used. It can be tricky to dumb the data down, intelligently. You may be needing to reduce 300 points in a 1x1m square, down to a single point. Generally, these complicated decisions will be streamlined through use of algorithms that can help smooth the overall data.
The data collected will have a number of attributes associated with it. One of the most important, will be the intensity reading from shots. This represents the reflectance of the object that the light bounced off of.
A higher intensity value indicates higher reflectance of the hit object.
Understanding the reflectance properties of objects in your collection area can help provide valuable context for this data, and allow the analyst to pick out specific types of targets.
One of the most important concepts for making point clouds usable is the ground truthing aspect. You have a point cloud, but where does it fit into space? Similar to georeferencing orthoimagery, you will need control points. Realistically, since the GNSS and IMU are generating high-quality spatial information, you really only need one control point for it to base its other calculations on. Since LiDAR is collecting elevation/surface data instead of images, this control point will need to be represented as an easily discernible object, so that it can later be located and used to reference in the point cloud.
In addition to control points, checkshots should also be collected.
These checkshots will be used to assess the accuracy of the resulting products which have been based on the drone’s data, and control point(s).
The analyst will be looking at absolute accuracy, relative accuracy, and point density to assess the success of the flight. Absolute accuracy is based on the control points, but the relative accuracy is based on comparing the returns from neighboring flight lines. This can help deduce outliers in the data, which can then be removed to improve the overall data quality.
So, when should a drone be considered for data collection? In the past, the only real options for aerial collection were planes and helicopters. Sometimes, these will still be the better option. Time, safety, and finances are going to be the biggest factors to consider here.
The upfront cost for a drone can be very expensive,
you may be looking at tens, to hundreds of thousands of dollars depending on the base drone itself, scanners, sensors, and secondary IMU and GNSS hardware
(what ships with the drone will likely not be at the caliber desired for reliable calculations). Once you have a drone, however, it is highly flexible. You now have all the diversity of movement that the pilot themselves can accomplish, via car, plane, and on foot, and the only additional costs, financial and time-wise, incurred are the movement of this individual. Planes and helicopters have a high initial cost to mobilize every time. Fuel and chartering costs can rack up quickly.
In terms of time, drones may not be the best option for collecting data for large swaths of land. At this point, it will be more efficient to scale up to a larger aerial vehicle that can collect larger footprints. Level of detail will be associated with the altitude the data was collected at. If you need highly detailed data for an area, a drone will be the better choice as you can collect as close to the ground as is safe and reasonable.
Drones have awesome capabilities, and so far, these have mostly been realized by the utility and engineering sectors. Engineers mostly value the highly detailed topographic information that can be collected. Ground surface models in particular are useful, as they can understand where the ground itself is, without any troublesome trees or buildings. The utility sector, however, has a lot of interest in these surface features. Power lines, for example, are a hot topic. Understanding how close vegetation comes to existing infrastructure can help assess maintenance needs more accurately, helping prevent fires resulting from downed limbs and the like.
As time goes on, and drone and LiDAR technology become more accessible, these uses will continue to explode and relate to more and more industries.
LiDAR has been around in one form or another since the 1960s. In the past 5 years or so, however, the technology has exploded as sensors and drones are getting smaller, lighter, and cheaper. This lowers the barrier for entry and is allowing more industries to discover how LiDAR and drones can help them.
As mentioned before, one of the greatest limitations for point clouds, is that existing software struggles to handle them at their full scale. This means that certain amounts of detail are lost to generalization. This will be a large area of improvement for the future. For now though, the data backlog will continue to grow.
One of the most exciting areas for future growth in the drone LiDAR collection arena is bathymetric collection. Underwater data can be collected by utilizing the “green laser” light, which lies on the 500 nanometer area of the spectrum. Quality of data collected is highly dependent on the water turbidity, as well as standard weather conditions you would consider before a land-based drone flight. One of the most popular applications for this data collection, is of shore lines by governments. This is good news, as it can come as a sign that climate change and sea-level rise are being considered as serious threats.
Crafting a quality application for a job you really want takes time, so you do not want to spread yourself too thin. When constructing your CV, it is important to keep your audience in mind. Realistically, the first set of eyes will likely be a computer algorithm, scraping the submitted CVs for certain keywords.